Why AI hasn’t touched the trades — and what could actually change that

This site's homepage makes a strong claim: licensed trades are structurally protected from automation. We stand by it — but a claim that strong deserves a serious stress test rather than a pep talk. Here's the honest version: what protects the trades, what's already changing, and the scenarios that would actually erode the moat.

The three moats, restated precisely

• Physical presence in unstructured environments. Robotics succeeds in structured, repeatable spaces — factory cells, warehouses. A 1962 crawlspace with improvised plumbing, a fire-damaged panel, an attic in August: these are the opposite of structured. The frontier of general-purpose robotics is still years from "diagnose and repair arbitrary faults in arbitrary buildings," and the economics are worse than the engineering.
• Personal liability. A licensed tradesperson signs permitted work under a personal license number, backed by insurance and a board that can revoke it. Liability needs a throat to choke; no manufacturer is volunteering to absorb open-ended liability for autonomous field work in homes.
• Statute. Who may legally perform electrical or plumbing work is written into state law. Changing it requires legislatures — fifty of them — to act against the interests of licensed constituencies. This is the slowest-moving moat, and arguably the strongest.

What AI is already changing (and it's real)

Pretending nothing changes would be its own dishonesty. Inside trade businesses, the office side is transforming quickly:

• Estimating and takeoffs: software now drafts bids from plans and photos in a fraction of the time — compressing one of the contractor's most labor-intensive office tasks.
• Scheduling, dispatch, and CRM: the service-business software stack increasingly optimizes routes, predicts job durations, and writes the follow-up emails.
• Diagnosis support: a tech with a multimeter and an AI assistant that has read every service manual is a faster tech. This augments field workers rather than replacing them — but it does compress the experience advantage of older techs.
• Design and code lookup: load calculations, code citations, and permit paperwork are exactly the kind of structured text work language models handle well.

Net effect so far: the business of the trades is automating; the work is not. That asymmetry favors tradespeople who also run businesses — one more argument for the contractor track we discuss in the career-switching article.

What could actually erode the moat

Intellectual honesty requires naming the real risks, roughly in order of plausibility:

1. New construction gets prefabricated. The genuine automation frontier isn't robots in crawlspaces — it's factories building wall panels, bathroom pods, and mechanical skids with their wiring and plumbing already installed in structured environments. Field trades don't disappear, but new-construction hours per building could shrink, shifting employment toward service, retrofit, and repair (which is most of the work anyway, and the least automatable part).
2. Licensing reform as deregulation. The moat is statutory, so the real threat is statutory: persistent labor shortages could pressure legislatures to widen handyman exemptions, raise unlicensed-work thresholds, or accept remote video inspection in place of licensed sign-off for more categories of work. Watch contractor-threshold bills, not robotics demos.
3. Liability innovation. If insurers ever build products that price autonomous field-work risk profitably, one pillar weakens. There is no sign of this today; insurers are conservative for good reasons.
4. Humanoid robotics actually delivering. The headline scenario and the least likely on any planning horizon that matters for a career decision today. Even granting the hardware, the permitting, liability, and consumer-trust layers would lag by many years.

The planning horizon answer

A 25-year-old entering an apprenticeship today will work until roughly 2070. No one can promise what 2070 looks like. What the structure supports promising: the moats protecting licensed field work are not the kind that erode quietly or quickly — they erode through visible, slow, political processes you can watch happening. Compare that to careers where the disruption arrives as a software update.